Key point driven spatial mode animal multi-level behavior analysis method and system

Abstract

The application provides a key point driven spatial mode animal multi-level behavior analysis method and system, and relates to the technical field of animal behavior analysis. The method comprises the following steps: in response to a first instruction, decoding and frame extraction processing are performed on experimental video data to generate key point time sequence trajectory data; based on a special motion mode detection algorithm matched with the type of the mode animal, analysis operation is performed in combination with the environment configuration parameters and the key point time sequence trajectory data to obtain a first analysis result; the first analysis result comprises an abnormal motion behavior type corresponding to the mode animal and a corresponding quantitative index, and the quantitative index is a characteristic parameter calculation value of the abnormal behavior. The application can realize automatic identification and quantitative analysis of abnormal motion behavior of the mode animal in a microgravity environment, provides core data support for mechanism exploration of the influence of microgravity on animal behavior in space biology research, and fills the gap of intelligent analysis tools for mode animal behavior in a space microgravity environment.

Description

Technical Field

[0001] This invention relates to the field of animal behavior analysis technology, and in particular to a key-point driven spatial model animal multi-level behavior analysis method and system. Background Technology

[0002] As on-orbit life science experiments continue to deepen, model organisms such as nematodes, zebrafish, and fruit flies have become core carriers for studying the effects of special environments such as microgravity and submagnetic fields on organisms, resulting in a continuous generation of a large number of on-orbit behavioral observation videos. The information contained in these video data, such as animal locomotion and abnormal behavioral patterns, is a key basis for revealing the mechanisms by which the space environment affects the physiological functions of organisms. Therefore, efficient and accurate behavioral analysis of these massive amounts of video data has become a core requirement to support in-depth data mining and scientific breakthroughs in space biology experiments.

[0003] Currently, there are two main types of technical solutions for analyzing animal behavior. One type relies on traditional methods that use manual annotation. Researchers observe video frame by frame, manually record the animal's movement trajectory, behavior type, and duration, and then perform statistical analysis based on the manually recorded data. The other type is automated analysis tools for ground-based scenarios. These tools typically integrate basic target detection and motion tracking algorithms, which can perform preliminary quantification of animal behavior in a controlled laboratory environment, such as determining whether an animal is moving or stationary by using a simple speed threshold.

[0004] However, existing technological solutions are insufficient to meet the specific needs of behavioral analysis of space-based model animals. Their core shortcomings lie in their lack of adaptability to the microgravity environment and their inability to perform cross-species analysis efficiently and accurately process the massive amounts of video data generated in space experiments and extract in-depth behavioral information. Manual annotation methods are not only extremely inefficient, unable to handle the terabytes of video data generated by space station experiments, but also suffer from strong subjectivity and poor consistency. Ground-based automated tools are mostly designed for the routine behavior of single species, failing to identify anomalous movement patterns unique to microgravity environments, such as zebrafish reversing their backs or fruit flies exhibiting malfunctioning movements. They also lack the ability to universally model behavioral characteristics across multiple species and have not established a standardized analytical loop from data preprocessing to result archiving. This makes it difficult to compare and reproduce experimental results from different batches and environments, severely hindering the progress of space life science research. Summary of the Invention

[0005] This invention provides a keypoint-driven multi-level behavior analysis method and system for space model animals, which can realize the automated identification and quantitative analysis of abnormal movement behavior of model animals under microgravity environment. It provides core data support for exploring the mechanism of microgravity's influence on animal behavior in space biology research and fills the gap in intelligent analysis tools for model animal behavior in space microgravity environment.

[0006] To achieve the above objectives, the embodiments of the present invention adopt the following technical solutions: Firstly, a keypoint-driven multi-level behavior analysis method for space model animals is provided. The method includes: receiving a first instruction input by a user, the first instruction instructing the identification and quantitative analysis of abnormal movement behavior of a space model animal under microgravity conditions, and the first instruction carrying model animal type, experimental video data, and environmental configuration parameters; responding to the first instruction, decoding and frame extraction processing of the experimental video data, generating keypoint temporal trajectory data based on the coordinates of feature keypoints of the model animal in each frame of the experimental video data; performing analysis operations based on a special movement pattern detection algorithm matching the model animal type, combined with the environmental configuration parameters and the keypoint temporal trajectory data, to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, the quantitative index being a calculated value of the feature parameters of the abnormal behavior.

[0007] The beneficial effects of this invention are as follows: The method provided by this invention receives a first instruction carrying the type of model animal, experimental video data, and environmental configuration parameters. Then, it decodes and extracts frames to generate key point temporal trajectory data. Finally, it matches the data with a special motion pattern detection algorithm to obtain analysis results containing abnormal behavior types and quantitative indicators. This enables automated identification and quantitative analysis of abnormal motion behavior in model animals under microgravity environments. The method provided by this invention not only eliminates the reliance on manual annotation and significantly improves the processing efficiency of massive amounts of video data, but also ensures the objectivity and reproducibility of the analysis results through animal type matching algorithms and quantitative indicator calculations. It provides core data support for exploring the mechanisms by which microgravity affects animal behavior in space biology research, filling the gap in intelligent analysis tools for model animal behavior in microgravity environments.

[0008] In one possible implementation of the first aspect, the model animal type includes zebrafish, fruit fly, and nematode. When the model animal type is zebrafish, the abnormal behavior type corresponding to the model animal includes spinning in place or turning belly-up. When the model animal type is fruit fly, the abnormal behavior type corresponding to the model animal includes deviation of velocity from posture direction, acceleration contrary to posture direction, or spinning in place. When the model animal type is nematode, the abnormal behavior type corresponding to the model animal includes head-tail asynchrony or disordered pulsation rhythm.

[0009] The method provided by this invention defines the corresponding abnormal behavior types for three typical model organisms: zebrafish, fruit flies, and nematodes (e.g., zebrafish circling and dorsal inversion, fruit flies exhibiting three types of movement failure, and nematodes exhibiting head-tail asynchrony and disordered pulsation rhythm), thus solving the problem of ambiguity in abnormal behavior types in cross-species behavioral analysis. By matching specific abnormal behavior types to different species, the method ensures the relevance and accuracy of the analysis, avoids analytical errors caused by confusion in behavior types, and lays a clear classification foundation for subsequent specialized analyses of the three species, supporting the development of cross-species spatial behavior comparative studies.

[0010] In one possible implementation of the first aspect, the special movement pattern detection algorithm based on the model animal type matching, combined with the environmental configuration parameters and key point temporal trajectory data, performs analysis operations to obtain a first analysis result, including: extracting the coordinates of key points on the zebrafish's mouth, back, abdomen, and tail based on the key point temporal trajectory data; determining the quantitative indicators corresponding to the zebrafish based on the environmental configuration parameters and the coordinates of the zebrafish's mouth, back, abdomen, and tail key points, the quantitative indicators including head-tail orientation angle, moving average angular velocity, and average translational velocity; determining the zebrafish's abnormal behavior type as circling in place when the zebrafish's moving average angular velocity is greater than the angular velocity threshold and the average translational velocity is less than the upper limit of the stationary velocity; and determining the zebrafish's abnormal behavior type as circling in place when the zebrafish's moving average angular velocity is greater than the angular velocity threshold and the average translational velocity is less than the upper limit of the stationary velocity. y Back,t Greater than y Abdomen,t In this case, the abnormal behavior of the zebrafish was determined to be dorsoventrally inverted. The formula for determining the head and tail orientation angle of a zebrafish in frame t is: θ t =arctan2(y Tail3,t −y Mouth,t ,x Tail3,t −x Mouth,t ) ; Among them, the coordinates of the key points of the zebrafish's mouth are ( x Mouth,t , y Mouth,t The coordinates of the key points on the back are ( x Back,t ， y Back,t The coordinates of the key points in the abdomen are ( x Abdomen,t , y Abdomen,t The coordinates of the key points at the tail are ( xTail3,t , y Tail3,t ); The moving average angular velocity of the zebrafish in frame t The formula for determining it is: ; ; Where W is the length of the sliding window. The wrap() function is used to reduce the angle difference between the head and tail orientation angles to the specified value. ; The formula for determining the average translational velocity of the zebrafish in frame t is: ; Where, p i Let be the center point of the zebrafish in the i-th frame.

[0011] The method provided by this invention extracts the coordinates of key points on the mouth, back, abdomen, and tail of zebrafish, calculates three types of quantitative indicators based on environmental configuration parameters: head-tail orientation angle, moving average angular velocity, and average translational velocity, and clarifies the criteria for determining stationary spinning and belly-up reversal. This achieves accurate identification and quantification of two unique abnormal behaviors in zebrafish. The method not only transforms abstract behavioral characteristics into calculable physical quantities through specific formulas—such as head-tail orientation angle reflecting direction changes and moving average angular velocity reflecting rotational intensity—but also avoids false positives such as spinning while walking or occasional posture deviations through strict judgment conditions. This provides a reliable quantitative basis for studying the mechanism of disorientation behavior in zebrafish under microgravity environments.

[0012] In one possible implementation of the first aspect, the special movement pattern detection algorithm based on the model animal type matching, combined with the environmental configuration parameters and key point temporal trajectory data, performs analysis operations to obtain a first analysis result, including: extracting the coordinates of key points on the mouth, left wing, right wing, back, abdomen, and tail of the fruit fly based on the key point temporal trajectory data; determining the quantitative indicators corresponding to the fruit fly based on the environmental configuration parameters and the coordinates of the key points on the mouth, left wing, right wing, back, abdomen, and tail of the fruit fly, the quantitative indicators including the velocity-attitude angle, the acceleration-attitude angle, and the attitude rotation angle; and determining the quantitative indicators corresponding to the velocity-attitude angle of the fruit fly based on the environmental configuration parameters and the coordinates of the key points on the mouth, left wing, right wing, back, abdomen, and tail of the fruit fly. If the angle between the attitude and attitude is greater than the direction deviation threshold, the magnitude of the velocity vector is greater than the first magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be a deviation between velocity and attitude direction; if the angle between the acceleration and attitude is greater than the acceleration contradiction threshold, the magnitude of the acceleration vector is greater than the third magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be an acceleration contradiction with the attitude direction; if the angle difference between the attitude rotation angles of the fruit fly is greater than the minimum rotational angular velocity threshold, and the magnitude of the velocity vector is less than the near-zero velocity threshold, the abnormal behavior type of the fruit fly is determined to be in-situ rotation. The velocity and attitude angle of the fruit fly in frame t The formula for determining it is: ; Among them, v t d is the velocity vector of the t-th frame; t The orientation is determined based on the coordinates of key points on the mouth and tail of the fruit fly; ∈[0,π]; The acceleration and attitude angle of a fruit fly in frame t The formula for determining it is: ; ; Among them, a t For acceleration; The formula for determining the attitude rotation angle of the fruit fly in frame t is: ; in, Let represent the attitude direction along the y-axis. The x-axis represents the attitude direction.

[0013] The method provided by this invention solves the problem of accurately identifying the types of motion failures in fruit flies. The method ensures the accuracy of attitude direction calculation through complete key point extraction, establishes the correlation between motion and attitude by quantifying the velocity-attitude angle and acceleration-attitude angle, and ultimately achieves a clear distinction between the three types of motion failures. This provides refined behavioral data for studying anomalies in the motion control mechanism of fruit flies under microgravity, improving the scientific rigor and reproducibility of the analysis results.

[0014] In one possible implementation of the first aspect, the special movement pattern detection algorithm based on the model animal type matching, combined with the environmental configuration parameters and key point temporal trajectory data, performs analysis operations to obtain a first analysis result, including: extracting the coordinates of 5 key points of the nematode based on the key point temporal trajectory data; determining the quantitative indicators corresponding to the nematode based on the environmental configuration parameters and the coordinates of the 5 key points of the nematode, the quantitative indicators including head swaying frequency, tail swaying frequency, and body undulation angle; determining the abnormal behavior type of the nematode as head-tail asynchrony when the absolute value of the difference between the head swaying frequency and the tail swaying frequency is greater than a frequency threshold; determining the swaying duration and swaying amplitude of every two adjacent swaying cycles of the nematode based on the body undulation angle of the nematode in multiple consecutive frames; determining the abnormal behavior type of the nematode as pulsation rhythm disorder when the difference between any two adjacent swaying durations exceeds a preset duration threshold or the difference between the swaying amplitudes exceeds a preset amplitude threshold; The formulas for determining the head swaying angle, tail swaying angle, and body undulation angle of the nematode in frame t are as follows: ; ; ; ; in, The angle between the two vectors, Let K be the coordinates of the k-th keypoint in frame t. The formula for determining the head-wagging frequency or tail-wagging frequency of a nematode is: ; Where C is the effective number of swings of the head or tail, the effective swing angle is greater than or equal to the preset angle threshold, and the number of continuous frames is greater than the preset number of continuous frames, T is the total number of frames, and F is the frame rate.

[0015] This invention provides a method that extracts five key feature points from nematodes, calculates three core indicators—head swaying angle, tail swaying angle, and body undulation angle—determines head-tail asynchrony based on swaying frequency differences, and identifies pulsation rhythm disorder based on differences in the duration and amplitude of adjacent swaying cycles. This achieves precise quantitative analysis of two types of subtle abnormal behaviors in nematodes. The method transforms the coordinated pulsation characteristics of nematodes into calculable angle and frequency parameters, and captures subtle changes in rhythm disorder through periodic stability analysis. This solves the problem of traditional methods' difficulty in quantifying subtle behavioral anomalies, providing crucial data support for studying the influence of microgravity on the circadian rhythm regulation mechanism of nematodes.

[0016] In one possible implementation of the first aspect, the method further includes: receiving a second instruction input by a user, the second instruction being used to instruct the quantitative analysis of low-order kinematic characteristics of a space model animal, and the second instruction carrying experimental video data and environmental configuration parameters; responding to the second instruction, decoding and frame-sampling processing of the experimental video data to obtain a skeletal center coordinate sequence of the model animal; and determining a second analysis result of the model animal based on the skeletal center coordinate sequence of the model animal, the second analysis result including velocity, acceleration, local average velocity, global average velocity, and activity. The velocity v t The formula for determining it is: ; in, Let t be the coordinates of the skeleton center. This refers to the inter-frame time interval. The acceleration a t The formula for determining it is: ; The local average velocity The formula for determining it is: ; The overall average speed The formula for determining it is: ; in, For the set of motion frame indices, ; Represents the Euclidean norm; k is the half-width of the window; The activity The formula for determining it is: ; Where T is the total number of frames; For indicator functions; This is the speed threshold.

[0017] The method provided by this invention receives a second instruction, then decodes and extracts frames to obtain the skeletal center coordinate sequence, and then calculates five low-order kinematic indices: velocity, acceleration, local average velocity, global average velocity, and activity, filling the gap in specialized analysis of basic kinematic characteristics. This method not only provides a basic reference for abnormal behavior analysis but also ensures the reproducibility of basic parameters through a unified calculation formula, forming a complete analytical chain of basic kinematic characteristics and abnormal behavior, providing data support for comprehensively characterizing the evolution of animal movement states under microgravity.

[0018] In one possible implementation of the first aspect, when the model animal is a fruit fly, the method further includes: receiving a third instruction input by a user, the third instruction being used to instruct the identification and statistical analysis of semantic-level behavioral features of the fruit fly, and the instruction carrying experimental video data; in response to the third instruction, extracting the head center coordinate sequence and feature key point angle data of the fruit fly, and determining the inter-frame displacement and total length of the historical trajectory of the fruit fly; determining the semantic-level behavioral category corresponding to each frame image based on the inter-frame displacement and total length of the historical trajectory of the fruit fly, the semantic-level behavioral category including stationary behavior, crawling behavior, flight behavior, courtship behavior, and mating behavior, wherein, when the inter-frame displacement of the fruit fly is less than the upper limit of the frame-level displacement and the total length of the historical trajectory is less than the short window cumulative path... Under certain conditions, the semantic-level behavior category is determined as stationary behavior. If the inter-frame displacement of the fruit fly is greater than or equal to the upper limit of frame-level displacement and the total length of the historical trajectory is greater than or equal to the upper limit of the cumulative path of the short window, the semantic-level behavior category is determined as crawling behavior. If the inter-frame displacement of the fruit fly is greater than or equal to the flight speed threshold, the semantic-level behavior category is determined as flight behavior. If the absolute value of the difference between the angles of the left and right wings relative to the trunk axis is greater than the left and right wing difference threshold, and the maximum value of the angles of the left and right wings relative to the trunk axis is greater than the minimum angle threshold, the semantic-level behavior category is determined as courtship behavior. If the distance between the centers of the abdomen or trunk of two fruit flies is less than the contact distance threshold, the orientation difference is less than the orientation difference threshold, and the number of consecutive frames is greater than the number of consecutive frames threshold, the semantic-level behavior category is determined as mating behavior.

[0019] The method provided by this invention receives a third instruction, extracts the head coordinates and wing angle data of a fruit fly, calculates the inter-frame displacement and the total length of the historical trajectory, and defines the judgment conditions for five semantic-level behaviors: stillness, crawling, flight, courtship, and mating. This enables the accurate identification of complex semantic-level behaviors in fruit flies. The method transforms abstract semantic behaviors into quantifiable parameters and avoids misjudgments caused by single parameters through the coordinated determination of inter-frame displacement and the total length of the historical trajectory. This provides semantic-level data support for studying the changing mechanisms of fruit fly social behavior under microgravity and expands the depth of space animal behavior analysis.

[0020] In one possible implementation of the first aspect, the method further includes: receiving a fourth instruction input by a user, the fourth instruction being used to instruct differential analysis of behavioral data from different environments or generations, and the instruction carrying at least two sets of experimental data, comparison dimension parameters, and significance test types; responding to the fourth instruction, extracting core quantitative indicators from each set of experimental data, the core quantitative indicators including abnormal movement behavior characteristics, low-order kinematic characteristics, and semantic-level behavioral characteristics of model animals; if the number of comparison groups K=2 and satisfies normality and homogeneity of variance, using a t-test to calculate the test statistic; if the number of comparison groups K=2 and does not satisfy normality or homogeneity of variance, using a Mann-Whitney U test to calculate the test statistic; if the number of comparison groups K>2, using one-way ANOVA to calculate the test statistic; if considering the interaction effect of two factors, using a linear model to calculate the test statistic; and outputting a fourth analysis result based on the test statistic, including inter-group difference values, significance determination results, and visual comparison charts.

[0021] The method provided by this invention receives a fourth instruction, extracts core quantitative indicators, and calculates test statistics using t-tests, Mann-Whitney U tests, ANOVA, or linear models based on the number of comparison groups and data characteristics. Finally, it outputs results including inter-group difference values, significance determination, and visualization charts, solving the problem of insufficient scientific statistical support for multi-factor behavioral data comparisons. By adapting statistical testing methods to different data characteristics, the method provides quantification and significance determination of inter-group differences (such as analyzing the interaction effect between environment and genotype using linear models), providing rigorous statistical evidence for the study of the mechanisms by which space environments (such as microgravity and submagnetic fields) affect animal behavior, and enhancing the scientific rigor and persuasiveness of research conclusions.

[0022] Secondly, the present invention provides a keypoint-driven multi-level behavior analysis system for space model animals. The system includes: an instruction receiving module for receiving a first instruction input by a user, the first instruction instructing the identification and quantitative analysis of abnormal movement behavior of space model animals in a microgravity environment, and the first instruction carrying model animal type, experimental video data, and environmental configuration parameters; a data processing module for responding to the first instruction, decoding and frame extraction of the experimental video data, and generating keypoint temporal trajectory data based on the coordinates of feature keypoints of the model animal in each frame of the experimental video data; and a detection module for performing analysis operations based on a special movement pattern detection algorithm matching the model animal type, combined with the environmental configuration parameters and the keypoint temporal trajectory data, to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, the quantitative index being a calculated value of the feature parameters of the abnormal behavior.

[0023] Thirdly, an electronic device is provided, the electronic device including a memory and one or more processors; the memory is coupled to the processors; wherein the memory stores computer program code, the computer program code including computer instructions, which, when executed by the processor, cause the electronic device to perform the method as described in any implementation of the first aspect.

[0024] Fourthly, a computer-readable storage medium is provided, including computer instructions that, when executed on an electronic device, cause the electronic device to perform a method as described in any implementation of the first aspect.

[0025] Fifthly, a computer program product is provided that, when run on a computer, causes the computer to perform the method in any implementation of the first aspect.

[0026] Understandably, the beneficial effects achieved by the analysis system of the second aspect, the electronic device of the third aspect, the computer-readable storage medium of the fourth aspect, and the computer program product of the fifth aspect provided above can be referred to with reference to the beneficial effects of the first aspect and any of its possible design embodiments, which will not be repeated here. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; Figure 2 A schematic diagram of the software structure of an analysis system provided in an embodiment of the present invention; Figure 3 A schematic diagram of the operation mechanism of an analysis system provided in an embodiment of the present invention; Figure 4 A flowchart illustrating a keypoint-driven, multi-level behavioral analysis method for spatial model animals, provided as an embodiment of the present invention; Figure 5 A flowchart of another keypoint-driven spatial model animal multilevel behavior analysis method provided in an embodiment of the present invention; Figure 6 A flowchart of yet another keypoint-driven spatial model animal multilevel behavior analysis method provided in an embodiment of the present invention; Figure 7 A flowchart of yet another keypoint-driven spatial model animal multilevel behavior analysis method provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the interface of a nematode pose estimation and behavior analysis platform provided in an embodiment of the present invention; Figure 9 A graph showing the difference in behavioral indicators of nematodes over time in ground and orbital environments, provided as an embodiment of the present invention; Figure 10 This is a diagram illustrating the behavioral differences between different strains of nematodes in terrestrial and orbital environments, provided as an embodiment of the present invention. Figure 11 An overall analysis result diagram provided for an embodiment of the present invention; Figure 12 This is a schematic diagram illustrating a multi-dimensional visualization result provided in an embodiment of the present invention; Figure 13 This is a schematic diagram of the structure of an analysis system provided in an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be described below with reference to the accompanying drawings. In the description of the present invention, unless otherwise stated, " / " indicates that the objects before and after are in an "or" relationship. For example, A / B can represent A or B. The "or" in the present invention is merely a description of the relationship between the related objects, indicating that three relationships can exist. For example, A or B can represent: A alone, A and B simultaneously, and B alone. A and B can be singular or plural. Furthermore, in the description of the present invention, unless otherwise stated, "multiple" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items.

[0029] Furthermore, to facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and that the terms "first" and "second" are not necessarily different.

[0030] In this embodiment of the invention, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in this embodiment of the invention should not be construed as superior or more advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0031] As on-orbit life science experiments continue to deepen, model organisms such as nematodes, zebrafish, and fruit flies have become core carriers for studying the effects of special environments such as microgravity and submagnetic fields on organisms, resulting in a continuous generation of a large number of on-orbit behavioral observation videos. The information contained in these video data, such as animal locomotion and abnormal behavioral patterns, is a key basis for revealing the mechanisms by which the space environment affects the physiological functions of organisms. Therefore, efficient and accurate behavioral analysis of these massive amounts of video data has become a core requirement to support in-depth data mining and scientific breakthroughs in space biology experiments.

[0032] Currently, there are two main types of technical solutions for analyzing animal behavior. One type relies on traditional methods that use manual annotation. Researchers observe video frame by frame, manually record the animal's movement trajectory, behavior type, and duration, and then perform statistical analysis based on the manually recorded data. The other type is automated analysis tools for ground-based scenarios. These tools typically integrate basic target detection and motion tracking algorithms, which can perform preliminary quantification of animal behavior in a controlled laboratory environment, such as determining whether an animal is moving or stationary by using a simple speed threshold.

[0033] However, existing technological solutions are insufficient to meet the specific needs of behavioral analysis of space-based model animals. Their core shortcomings lie in their lack of adaptability to the microgravity environment and their inability to perform cross-species analysis efficiently and accurately process the massive amounts of video data generated in space experiments and extract in-depth behavioral information. Manual annotation methods are not only extremely inefficient, unable to handle the terabytes of video data generated by space station experiments, but also suffer from strong subjectivity and poor consistency. Ground-based automated tools are mostly designed for the routine behavior of single species, failing to identify anomalous movement patterns unique to microgravity environments, such as zebrafish reversing their backs or fruit flies exhibiting malfunctioning movements. They also lack the ability to universally model behavioral characteristics across multiple species and have not established a standardized analytical loop from data preprocessing to result archiving. This makes it difficult to compare and reproduce experimental results from different batches and environments, severely hindering the progress of space life science research.

[0034] In view of this, embodiments of the present invention provide a keypoint-driven multi-level behavior analysis method and system for space model animals. The method includes: receiving a first instruction input by a user, the first instruction being used to instruct the identification and quantitative analysis of abnormal movement behavior of space model animals in a microgravity environment, and the first instruction carrying model animal type, experimental video data, and environmental configuration parameters; responding to the first instruction, decoding and frame extraction processing of the experimental video data, generating keypoint temporal trajectory data based on the coordinates of feature keypoints of the model animal in each frame of the experimental video data; performing analysis operations based on a special movement pattern detection algorithm matching the model animal type, combined with the environmental configuration parameters and the keypoint temporal trajectory data, to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, the quantitative index being a calculated value of the feature parameters of the abnormal behavior.

[0035] The beneficial effects of this invention are as follows: The method provided by this invention receives a first instruction carrying the type of model animal, experimental video data, and environmental configuration parameters. Then, it decodes and extracts frames to generate key point temporal trajectory data. Finally, it matches the data with a special motion pattern detection algorithm to obtain analysis results containing abnormal behavior types and quantitative indicators. This enables automated identification and quantitative analysis of abnormal motion behavior in model animals under microgravity environments. The method provided by this invention not only eliminates the reliance on manual annotation and significantly improves the processing efficiency of massive amounts of video data, but also ensures the objectivity and reproducibility of the analysis results through animal type matching algorithms and quantitative indicator calculations. It provides core data support for exploring the mechanisms by which microgravity affects animal behavior in space biology research, filling the gap in intelligent analysis tools for model animal behavior in microgravity environments.

[0036] In some embodiments, a keypoint-driven multilevel behavior analysis method for spatial model animals provided by the present invention can be executed by a keypoint-driven multilevel behavior analysis system 100 for spatial model animals (hereinafter referred to as analysis system 100).

[0037] As an example, the analysis system 100 can be any electronic device 200 with data processing capabilities, such as a general-purpose computer, personal computer, laptop computer, switch, or tablet computer. The specific implementation of the analysis system 100 is not limited here.

[0038] Figure 1 A schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present invention is shown. The electronic device 200 includes a processor 210, a memory 220, and a communication interface 230.

[0039] Processor 210 may include one or more processing cores. Processor 210 connects to various parts within electronic device 200 using various interfaces and lines, and performs various functions and processes data of electronic device 200 by running or executing instructions, programs, code sets, or instruction sets stored in memory 220, and by calling data stored in memory 220. Optionally, processor 210 may be implemented using at least one of the following hardware forms: Central Processing Unit (CPU), Graphics Processing Unit (GPU), Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA).

[0040] The memory 220 may include random access memory (RAI) or read-only memory (ROI). Optionally, the memory 220 may include non-transitory computer-readable storage ledger. The memory 220 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 220 may include a stored program area. The stored program area may store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the various method embodiments described above, etc.

[0041] Communication interface 230 is used to communicate with other devices, equipment or communication networks, such as data storage devices, image processing devices or Ethernet, wireless access network (RAN), wireless local area network (WLAN), etc.

[0042] In terms of physical implementation, the aforementioned devices (such as processor 210, memory 220, and communication interface 230) can each be devices within the same device (such as a laptop computer). Alternatively, at least two of these devices can be located within the same device, i.e., as different devices within the same device, similar to the deployment of devices or components in a distributed system.

[0043] It is understood that the structure illustrated in this embodiment does not constitute a specific limitation on the electronic device 200. In other embodiments of the present invention, the electronic device 200 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0044] In one example, see Figure 2 The analysis system 100 adopts a three-layer structure: the user interface layer (UI), the system function layer (Service), and the data management layer (Data). These three layers collaborate through standardized interface contracts: providing unified interaction and interpretable visualization upwards, and ensuring data consistency, traceability, and reproducibility downwards. The overall principles are as follows: Research task-driven: supporting continuous modeling and quantitative analysis of model animals such as fruit flies, nematodes, and zebrafish in terrestrial / in-orbit and geomagnetic / submagnetic environments; Layered decoupling and pluggable: clear division of responsibilities for the interface, algorithms, and data; detection / attitude / tracking / identification / analysis modules are all encapsulated as plug-ins for easy replacement and expansion; Structured output: uniformly producing behavior matrices, duration statistics, transition matrices, n-gram counts, behavioral spectra, and visualization artifacts, directly serving subsequent scientific research; Reproducibility and comparability: standardized data flow, containerized environment, task configuration, and versioned metadata support cross-session and cross-batch comparisons.

[0045] Specifically, the user interface (UI) layer serves as a one-stop interactive portal for researchers, used for project creation, data import / frame extraction, parameter configuration, training / inference monitoring, deep analysis, and report export. It implements a cross-platform desktop interface using Electron + HTML / CSS / JavaScript; it communicates with the backend via HTTP / IPC to transmit JSON messages (task submission, status subscription, result retrieval), and supports interactive displays such as keypoint / box overlay, trajectory visualization, frame-by-frame behavior sequences, transition heatmaps, and multi-column Sankey diagrams, as well as batch export.

[0046] The system functional layer (Service) is responsible for the main algorithms and analysis logic: video decoding and frame extraction, detection, pose estimation, tracking, behavior recognition, statistical analysis, and visualization rendering. The service framework is based on Python + Flask / FastAPI; task orchestration and asynchronous scheduling use Celery / RQ, supporting local GPU workstations and remote inference nodes; the model runtime environment is solidified using Docker containers (CUDA / cuDNN, PyTorch, NumPy, etc.) to ensure cross-machine consistency. The functional layer exposes REST / IPC interfaces to the upper layer and provides unified read / write access to the data layer; it has built-in runtime logs and metric collection (Precision / Recall, duration error, throughput / latency), and provides intermediate result feedback for manual review and parameter fine-tuning.

[0047] The Data Management Layer (Data) is used to uniformly manage raw videos / frames, annotation and inference results, statistical outputs, visualization artifacts, and model weights and runtime metadata (hyperparameters, timestamps, device information). The underlying storage defaults to SQLite (lightweight deployment), which can be smoothly switched to PostgreSQL in multi-user / high-concurrency scenarios. Data is organized around unified primary keys: video_id, frame_id, and track_id. Annotation and inference results are stored using a hybrid of COCO-style JSON and structured tables, ensuring that structured outputs can be stably indexed, recalculated, and compared. Paths and naming conventions are archived hierarchically according to "Experiment Number—Species—Date / Batch—Environmental Conditions".

[0048] It should also be noted that the analysis system 100 uses standardized JSON as the cross-module data contract, running through data collection, algorithm reasoning, statistical analysis, and visualization to form an automated closed-loop process. The operating mechanism is as follows: Figure 3 As shown: Project creation and data import: Configure information such as video_id, species, environment, and frame rate in the UI, supporting local / cloud video import; Frame extraction and annotation: The functional layer performs decoding and frame extraction, supporting multi-task annotation of key points, bounding boxes, individual IDs, and behavior categories; Provides manual review and re-annotation; Model training and inference: Generate training sets from annotations, and perform detection / pose / tracking / recognition; Inference results (key points, track_id, behavior labels, and scores) are written back to the data layer; Deep behavior analysis and visualization: Calculate activity indicators (velocity, acceleration, displacement, single duration), special behaviors (fruit fly movement failure, zebrafish disorientation, nematode swarming frequency), and high-order modeling (transition matrix, 𝑛-gram behavior spectrum, Sankey diagram) based on inference results; Export and archiving: One-click export of CSV / JSON and PNG / PDF, together with container images and parameter lists, forming a reproducible experimental closed loop of "data—code—configuration—output".

[0049] In summary, the analysis system 100 provided in this embodiment of the invention not only achieves a fully automated closed loop from data import and model inference to result export at the engineering operation level, but also realizes the identification of multi-scale activity curves, abnormal motion patterns, and higher-order behavioral units above attitude and tracking output, supporting systematic comparative analysis across individuals, days, generations, and environments. Meanwhile, thanks to its plug-in and interface-based design, the analysis system 100 can seamlessly integrate large models / intelligent agents (such as language-guided behavior retrieval, cross-modal semantic alignment, and automated evidence chain interpretation) in the future, further enhancing the intelligence and platform capabilities of on-orbit behavior analysis.

[0050] The following description, in conjunction with the accompanying drawings, illustrates a keypoint-driven, multi-level behavioral analysis method for spatial model animals provided by an embodiment of the present invention.

[0051] Figure 4 This is a flowchart illustrating a keypoint-driven, spatial model-based, multi-level behavioral analysis method for animals, provided as an embodiment of the present invention. Optionally, this method can be... Figure 1 The illustrated electronic device 200 performs this operation. The method may include the following steps: S1. Receive a first instruction input by the user. The first instruction is used to instruct the identification and quantitative analysis of abnormal movement behavior of space model animals in a microgravity environment. The first instruction carries the model animal type, experimental video data and environmental configuration parameters. S2. In response to the first instruction, the experimental video data is decoded and frame-stripped, and key point temporal trajectory data is generated based on the coordinates of the feature key points of the model animal in each frame of the experimental video data. S3. Based on the special movement pattern detection algorithm matching the model animal type, and combined with the environmental configuration parameters and key point time-series trajectory data, perform analysis and calculation to obtain the first analysis result.

[0052] The first analysis result includes the abnormal movement behavior type and corresponding quantitative index of the model animal, wherein the quantitative index is the calculated value of the characteristic parameter of the abnormal behavior.

[0053] In one possible implementation, the model animal types include zebrafish, fruit flies, and nematodes.

[0054] In some embodiments, when the model animal is zebrafish, the abnormal behavior type corresponding to the model animal includes spinning in place or turning upside down. Further, see Figure 5 The above S3 includes: S311. Extract the coordinates of key points of the zebrafish's mouth, back, abdomen and tail based on the key point time-series trajectory data; S312. Determine the quantitative indicators corresponding to the zebrafish based on the environmental configuration parameters and the coordinates of key points of the zebrafish's mouth, back, abdomen and tail. The quantitative indicators include the head and tail orientation angle, the moving average angular velocity and the average translational velocity. Specifically, the formula for determining the head and tail orientation angle of the zebrafish in frame t is: θ t =arctan2(y Tail3,t −y Mouth,t ,x Tail3,t −x Mouth,t ) ; Among them, the coordinates of the key points of the zebrafish's mouth are ( x Mouth,t , y Mouth,t The coordinates of the key points on the back are ( x Back,t ， y Back,t The coordinates of the key points in the abdomen are ( x Abdomen,t , y Abdomen,t The coordinates of the key points at the tail are ( x Tail3,t , y Tail3,t ); The moving average angular velocity of the zebrafish in frame t The formula for determining it is: ; ; Where W is the length of the sliding window. The wrap() function is used to reduce the angle difference between the head and tail orientation angles to the specified value. ; The formula for determining the average translational velocity of the zebrafish in frame t is: ; Where, p i Let be the center point of the zebrafish in the i-th frame.

[0055] S313. When the zebrafish's sliding average angular velocity is greater than the angular velocity threshold and the average translational velocity is less than the upper limit of the stationary velocity, the abnormal behavior type of the zebrafish is determined to be spinning in place.

[0056] Specifically, the determination of spinning in place is as follows: ; In the formula: Angular velocity threshold; v is the average translational velocity within W frames; spin The maximum speed in place (to suppress false positives in "turning and moving"); p t The center point of the zebrafish (e.g., the midpoint of Mouth-Tail3).

[0057] S314, In zebrafish y Back,t Greater than y Abdomen,t In this case, the abnormal behavior of the zebrafish was determined to be dorsoventrally inverted. Specifically, the determination of an inverted abdomen and back is as follows: ; In the formula: the y-axis is downward by default; "Back" is above the image and "Abdomen" is below. If the camera orientation or sample is flipped, a vector should be used instead. Make a consistency determination with the reference direction (such as the direction of gravity or the camera calibration normal).

[0058] As described in S311-S314 above, the method provided in this embodiment of the invention achieves accurate identification and quantification of two unique abnormal behaviors of zebrafish by extracting the coordinates of key points on the mouth, back, abdomen, and tail of the zebrafish, calculating three types of quantitative indicators based on environmental configuration parameters: head-tail orientation angle, moving average angular velocity, and average translational velocity, and clarifying the judgment conditions for stationary spinning and belly-back reversal. The method provided by this invention not only transforms abstract behavioral characteristics into calculable physical quantities through specific formulas, such as the head-tail orientation angle reflecting direction changes and the moving average angular velocity reflecting rotation intensity, but also avoids false positives such as spinning while walking and occasional posture deviations through strict judgment conditions, providing a reliable quantitative basis for the study of the mechanism of disorientation behavior in zebrafish under microgravity environments.

[0059] In other embodiments, when the model animal is fruit fly, the abnormal behavior types corresponding to the model animal include deviation of speed from posture direction, acceleration contrary to posture direction, or rotation in place. Further, see Figure 6 The above S3 includes: S321. Extract the coordinates of key points of the fruit fly's mouth, left wing, right wing, back, abdomen and tail based on the key point time-series trajectory data; S322. Based on the environmental configuration parameters and the coordinates of key points of the fruit fly's mouth, left wing, right wing, back, abdomen, and tail, determine the quantitative indicators corresponding to the fruit fly. The quantitative indicators include the angle between velocity and attitude, the angle between acceleration and attitude, and the attitude rotation angle. Specifically, the angle between the velocity and attitude of the fruit fly in frame t. The formula for determining it is: ; Among them, v t d is the velocity vector of the t-th frame; t The orientation is determined based on the coordinates of key points on the mouth and tail of the fruit fly; ∈[0,π], in the implementation, the ratio should be truncated using clip(·,-1,1) and ||v|| t ‖,‖d t Set a minimum threshold ε to avoid numerical instability.

[0060] The acceleration and attitude angle of a fruit fly in frame t The formula for determining it is: ; ; Among them, a t For acceleration, it is necessary to adjust ||a|| t Let the minimum threshold ε be... a And create a clip. The formula for determining the attitude rotation angle of the fruit fly in frame t is: ; in, Let represent the attitude direction along the y-axis. The x-axis represents the attitude direction.

[0061] S323. When the angle between the speed and attitude of the fruit fly is greater than the direction deviation threshold, the magnitude of the speed vector is greater than the first magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be a deviation between speed and attitude direction. Specifically, the deviation between velocity and attitude direction is determined as follows: ; In the formula: For directional deviation from the threshold; δ v ,,δ d These are the first modulus threshold and the second modulus threshold, respectively.

[0062] S324. When the angle between the acceleration and attitude of the fruit fly is greater than the acceleration contradiction threshold, the magnitude of the acceleration vector is greater than the third magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be acceleration and attitude direction contradiction. Specifically, the determination that acceleration and attitude direction are contradictory is as follows: ; In the formula: For accelerations that deviate from the threshold (such as large angles close to π).

[0063] S325. When the angle difference of the fruit fly's posture rotation angle is greater than the minimum rotation angular velocity threshold and the magnitude of the velocity vector is less than the near-zero velocity threshold, the abnormal behavior type of the fruit fly is determined to be in-situ rotation. Specifically, the determination of rotation in place is as follows: ; In the formula: δ v The near-zero velocity threshold; δ r The minimum rotational angular velocity threshold (radians / frame) can be used as an additional criterion to reduce interference from local movements such as "wing combing".

[0064] As described in S321-S325 above, the method provided by this invention solves the problem of accurate identification of Drosophila motion failure types. The method ensures the accuracy of attitude direction calculation through complete key point extraction, establishes the correlation between motion and attitude by quantifying velocity-attitude angle and acceleration-attitude angle, and ultimately achieves a clear distinction between the three types of motion failures. This provides refined behavioral data for studying anomalies in the motion control mechanism of Drosophila under microgravity, improving the scientific rigor and reproducibility of the analysis results.

[0065] In some other embodiments, when the model animal is a nematode, the abnormal behavior type corresponding to the model animal includes head-tail asynchrony or pulsation rhythm disorder.

[0066] Furthermore, participate Figure 7 The aforementioned S3 specifically includes: S331. Extract the coordinates of 5 key points of the nematode based on the key point time-series trajectory data; S332. Determine the quantitative indicators corresponding to the nematode based on the environmental configuration parameters and the coordinates of the 5 key points of the nematode. The quantitative indicators include the head swaying frequency, the tail swaying frequency, and the body undulation angle. Specifically, the formulas for determining the head swaying angle, tail swaying angle, and body undulation angle of the nematode in frame t are as follows: ; ; ; ; in, The angle between the two vectors, Let K be the coordinates of the k-th keypoint in frame t. The formula for determining the head-wagging frequency or tail-wagging frequency of a nematode is: ; Where C is the effective number of swings of the head or tail, the effective swing angle is greater than or equal to the preset angle threshold, and the number of continuous frames is greater than the preset number of continuous frames, T is the total number of frames, and F is the frame rate.

[0067] Specifically, for robustness, the ratios are cropped to [-1, 1] during calculation, and minimum thresholds are set for |a| and |b| to avoid numerical instability. For statistically effective oscillation, angular amplitude and shortest duration control parameters are defined, δ... angle (Minimum angular amplitude, default 20°), δlength (minimum duration of frames, default 2 frames).

[0068] S333. When the absolute value of the difference between the head swaying frequency and the tail swaying frequency is greater than the frequency threshold, the abnormal behavior type of the nematode is determined to be head-tail asynchrony. S334. Based on the body undulation angle of the nematode in consecutive frames, determine the undulation duration and amplitude of each two adjacent undulation cycles of the nematode. If the difference between any two adjacent undulation durations exceeds a preset duration threshold or the difference between the undulation amplitudes exceeds a preset amplitude threshold, determine the abnormal behavior type of the nematode as pulsating rhythm disorder. As described in S1-S3 above, the method provided by this invention receives a first instruction carrying the model animal type, experimental video data, and environmental configuration parameters. It then decodes and extracts frames to generate key point temporal trajectory data. Finally, it matches the data with a special motion pattern detection algorithm to obtain analysis results containing abnormal behavior types and quantitative indicators. This enables automated identification and quantitative analysis of abnormal motion behavior in model animals under microgravity conditions. The method not only eliminates the reliance on manual annotation and significantly improves the processing efficiency of massive amounts of video data, but also ensures the objectivity and reproducibility of the analysis results through animal type matching algorithms and quantitative indicator calculations. It provides core data support for exploring the mechanisms by which microgravity affects animal behavior in space biology research, filling the gap in intelligent analysis tools for model animal behavior in microgravity environments.

[0069] In some embodiments, the method provided by the present invention further includes: The system receives a second instruction input by the user, which instructs the quantitative analysis of low-order kinematic characteristics of a space model animal, and the second instruction carries experimental video data and environmental configuration parameters; in response to the second instruction, the system decodes and extracts frames from the experimental video data to obtain the skeletal center coordinate sequence of the model animal; based on the skeletal center coordinate sequence of the model animal, the system determines a second analysis result of the model animal, which includes velocity, acceleration, local average velocity, global average velocity, and activity.

[0070] Given a skeleton center sequence The speed v t The formula for determining it is: ; in, Let t be the coordinates of the skeleton center. This is the inter-frame time interval; for boundary processing, the velocity at t=1 can be set to missing / zero or extrapolated from the next frame.

[0071] The acceleration a t The formula for determining it is: ; Among them, a t This is the inter-frame acceleration vector, in pixels per second. 2 Its definition requires t greater than or equal to 3 to have sufficient history terms; accelerations with t less than 3 are treated as missing, 0, or steady-state assumptions. To improve robustness, v can be adjusted first. t Perform vector-level smoothing and then differencing.

[0072] For velocity smoothing and local averaging: To suppress measurement noise, a moving average is applied to the velocity modulus. The local average velocity is defined as follows: The local average velocity The formula for determining it is: ; The overall average speed The formula for determining it is: ; in, For the set of motion frame indices, ; This represents the Euclidean norm; k is the half-width of the window, and the set of motion frame indices is determined by the time window or frame-level threshold. The unit is pixels per second, reflecting the speed level when in motion. k=2 corresponds to a 5-frame window, with window shrinking or mirror padding used at the boundaries to avoid exceeding the limits.

[0073] The activity The formula for determining it is: ; Where T is the total number of frames; For indicator functions; This is the velocity threshold, measured in pixels per second, and can be a global or individual adaptive threshold (such as a quantile threshold). For noise reduction, [the following can be used]... replace Make a judgment.

[0074] In one example, the method provided in this embodiment of the invention responds to a second instruction, calculates low-order kinematic quantities such as velocity, acceleration, displacement, and duration of a single motion based on the trajectory of the center point of an individual skeleton, and supports activity analysis based on time windows to generate trend curves reflecting the evolution of the motion state. To reduce the high-frequency jitter caused by keypoint estimation offsets, a robust estimation strategy combining time window constraints and moving averages is adopted, as follows: First, time windows are divided and motion states are determined. Time windows (timebins) are divided with a fixed frame length (e.g., 25 frames), and the net displacement of the trajectory within each time window is counted. If the net displacement exceeds a threshold (default 3 pixels), it is recorded as "movement"; otherwise, it is recorded as "stationary". Based on this, the motion duration and stationary duration of an individual can be accumulated. To enhance robustness, the proportion of active frames within a window can be optionally counted as an auxiliary factor.

[0075] Secondly, the calculation of instantaneous kinematic parameters (velocity, acceleration, local average velocity, global average velocity, and mobility) is detailed in the description of the above embodiments and will not be repeated here.

[0076] In this way, the motion state curves and activity time series included in the second analysis output can serve as key quantitative evidence for assessing the microgravity adaptation process, and provide indirect and representative behavioral indicators for analyzing phenomena such as muscle atrophy, bone loss, and decreased motor ability in space experiments. If it is necessary to convert pixel dimensions into physical quantities (such as mm / s), a pixel-length scaling factor can be introduced in the experimental calibration.

[0077] As described above, the method provided by this invention solves the problem of accurately identifying the types of motion failures in fruit flies. The method ensures the accuracy of attitude direction calculation through complete key point extraction, establishes the correlation between motion and attitude by quantifying the velocity-attitude angle and acceleration-attitude angle, and ultimately achieves a clear distinction between the three types of motion failures. This provides refined behavioral data for studying anomalies in the motion control mechanism of fruit flies under microgravity, improving the scientific rigor and reproducibility of the analysis results.

[0078] As described above, by receiving the second instruction, decoding and extracting frames to obtain the skeleton center coordinate sequence, and then calculating five low-order kinematic indices—velocity, acceleration, local average velocity, global average velocity, and activity—this fills the gap in specialized analysis of basic kinematic characteristics. The method provided by this invention not only provides a basic reference for abnormal behavior analysis but also ensures the reproducibility of basic parameters through a unified calculation formula, forming a complete analytical chain of basic kinematic characteristics and abnormal behavior. This provides data support for comprehensively characterizing the evolution of animal movement states under microgravity.

[0079] In some other embodiments, when the model organism is fruit fly, the method further includes: The system receives a third instruction from the user, which instructs the identification and statistical analysis of semantic-level behavioral features of fruit flies, and the instruction carries experimental video data. In response to the third instruction, it extracts the head center coordinate sequence and feature key point angle data of the fruit fly, and determines the inter-frame displacement and total length of the historical trajectory of the fruit fly. Based on the inter-frame displacement and total length of the historical trajectory of the fruit fly, it determines the semantic-level behavioral category corresponding to each frame image. The semantic-level behavioral categories include stationary behavior, crawling behavior, flight behavior, courtship behavior, and mating behavior. Specifically, if the inter-frame displacement of the fruit fly is less than the upper limit of the frame-level displacement and the total length of the historical trajectory is less than the upper limit of the cumulative path of the short window, the semantic-level behavioral category is determined to be stationary behavior. When the inter-frame displacement of the fruit fly is greater than or equal to the upper limit of frame-level displacement and the total length of the historical trajectory is greater than or equal to the upper limit of the cumulative path of the short window, the semantic-level behavior category is determined to be crawling behavior; when the inter-frame displacement of the fruit fly is greater than or equal to the flight speed threshold, the semantic-level behavior category is determined to be flight behavior; when the absolute value of the difference between the angles of the left and right wings relative to the trunk axis of the fruit fly is greater than the left and right wing difference threshold, and the maximum value of the angles of the left and right wings relative to the trunk axis is greater than the minimum angle threshold, the semantic-level behavior category is determined to be courtship behavior; when the distance between the centers of the abdomen or trunk of two fruit flies is less than the contact distance threshold, the orientation difference is less than the orientation difference threshold, and the number of consecutive frames is greater than the number of consecutive frames threshold, the semantic-level behavior category is determined to be mating behavior.

[0080] The following example illustrates the process by which the analysis system responds to a third instruction and obtains a third analysis result.

[0081] Specifically, to fully characterize the animal behavior features in a microgravity environment, the present example constructs a hierarchical modeling framework from behavior units to an ethogram based on key point time-series trajectories and biology-inspired prior rules, automatically identifies typical and rare behaviors, and outputs structured labels, statistical maps and visualization results. The present method does not rely on training-based boundary detection or a fixed sliding window classifier, but combines physical quantities such as speed, displacement, and angle change to form a lightweight, interpretable and cross-species transferable high-order behavior analysis process. By default, angles are measured in radians, the video frame rate is F, and the frame interval Δt = 1 / F.

[0082] First, behavior rule modeling is performed (taking Drosophila as an example).

[0083] Let the coordinate of the head center at the t-th frame be P∈R 2 . Define: Inter-frame displacement: ; In the formula: d t is the Euclidean distance (in pixels) between the head centers of adjacent frames; p t =(x t , y t ) is the coordinate of the t-th frame. The position at t=1 can be set as missing or extrapolated.

[0084] Total historical trajectory length (past W frames): ; In the formula: L t is the accumulated path length (in pixels) within the time window; W is the window length; at the boundary (t<W), window shrinking or zero-padding processing is adopted.

[0085] Stationary behavior criterion: ; In the formula: δ rest is the upper limit of frame-level displacement (pixels / frame), δ length is the upper limit of the accumulated path in the short window (pixels), which jointly suppresses jitter false detection.

[0086] Crawling behavior criterion: ; In the formula: δ fry is the flight speed threshold (pixels / frame); adding the lower limit of L t ensures continuous displacement rather than occasional jitter.

[0087] Flight behavior criterion: ; In the formula: the 2D projection speed is used as an approximate indicator of flight; if there are three-dimensional cues or out-of-focus blur intensity, they can be used as auxiliary conditions.

[0088] Criteria for courtship behavior: ; In the formula: The angle (in radians) between the left and right wings relative to the trunk axis can be determined by... =∠(LeftWing t -Back t Calculated by Tailt-Moutgt, δ wing δ represents the threshold for difference between the left and right wings. spread It is the minimum opening angle.

[0089] Mating criteria (bibody contact + orientation alignment + duration): ; In the formula, The distance (in pixels) between the centers of the two abdomens or torsos; ; Dur i,j The number of consecutive frames that satisfy the first two conditions; δ mate δ angle and δ dur For contact distance, alignment angle and shortest duration threshold.

[0090] Next comes behavioral statistics and spectrogram construction. After obtaining frame-by-frame labels, structured statistics and visualization are performed to form a behavioral spectrogram.

[0091] The tag matrix is ​​as follows: ; In the formula: For frames The category matrix of individuals, where C is the set of behavioral categories.

[0092] Duration and frequency (by individual i, category c): ; In the formula: t k,* For the k-th segment of category c, the start and end frames are defined. # is the frame interval; #episodes; (c) is the number of segments.

[0093] Finally, the behavior transition matrix (row normalized, with Laplace smoothing) is: ; In the formula: C ab This is the frame transition count from a to b; a>0 is the smoothing coefficient (e.g., 10). -3 To avoid zero probability; .

[0094] The above statistical outputs correspond to the following spectral elements: behavior label matrix heatmap, duration / frequency distribution map, transition matrix heatmap, and Sankey diagram (from P). ab It can be mapped to the edge width and extended to the joint distribution of "behavior-space" (combined with ROI heatmap), providing interpretable support for the construction of staged behavior chains and the recognition of abnormal patterns.

[0095] As described above, the method provided by this invention solves the problem of accurately identifying the types of motion failures in fruit flies. The method ensures the accuracy of attitude direction calculation through complete key point extraction, establishes the correlation between motion and attitude by quantifying the velocity-attitude angle and acceleration-attitude angle, and ultimately achieves a clear distinction between the three types of motion failures. This provides refined behavioral data for studying anomalies in the motion control mechanism of fruit flies under microgravity, improving the scientific rigor and reproducibility of the analysis results.

[0096] In some embodiments, the method further includes: receiving a fourth instruction input by a user, the fourth instruction being used to instruct differential analysis of behavioral data from different environments or generations, and the instruction carrying at least two sets of experimental data, comparison dimension parameters, and significance test types; responding to the fourth instruction, extracting core quantitative indicators from each set of experimental data, the core quantitative indicators including abnormal movement behavior characteristics, low-order kinematic characteristics, and semantic-level behavioral characteristics of model animals; if the number of comparison groups K=2 and satisfies normality and homogeneity of variance, using a t-test to calculate the test statistic; if the number of comparison groups K=2 and does not satisfy normality or homogeneity of variance, using a Mann-Whitney U test to calculate the test statistic; if the number of comparison groups K>2, using one-way ANOVA to calculate the test statistic; if considering the interaction effect of two factors, using a linear model to calculate the test statistic; and outputting a fourth analysis result including inter-group difference values, significance determination results, and visual comparison charts based on the test statistic.

[0097] Specifically, the fourth instruction mentioned above can be understood as being used to assess the impact of different experimental conditions and individual attributes on behavioral indicators, in order to support grouping and significance testing across dimensions such as environment (ground / on-orbit, geomagnetic / submagnetic, intravehicular / extravehicular radiation), time scale (hourly / daily), generation, and genotype. This module takes the individual / session-level indicators derived from the previous section as input (such as duration, frequency, activity, number of failure events, probability of behavioral transition, etc.) and outputs grouped statistical tables, significance markers, and visualization maps.

[0098] In one example, in response to the fourth instruction, the analysis system first performs indicator aggregation and grouping, denoting the observation of a certain behavioral indicator at the individual (or conversation) level as {M}. u}, where u represents a unique individual-conversation unit. Let the grouping factor be G∈{g,…,gk} (e.g., environment, genotype, or their Cartesian product). The within-group mean is defined as: ; in, This is the set of sample indices belonging to group g. This is the sample size for this group.

[0099] Furthermore, for comparisons that include time factors (by hour / by day), the observations are first aggregated by time bucket b∈B; ; in, This represents the set of samples that simultaneously satisfy both group g and time bucket b. Its sample size.

[0100] Then, comparisons are made. When comparing two independent groups, if they approximately satisfy normality and homogeneity of variance, a test is used; otherwise, the Mann-Whitney U test (rank-sum test) is used. For paired data, either a paired test or the Wilcoxon signed-rank test is used. When comparing multiple independent groups, one-way ANOVA or the Kruskal-Wallis test can be used; if the interaction effect of two factors (such as environment and genotype) is considered simultaneously, a linear model is used.

[0101] The linear model is: ; The method provided in this invention receives a fourth instruction, extracts core quantitative indicators, and calculates test statistics using t-tests, Mann-Whitney U tests, ANOVA, or linear models based on the number of comparison groups and data characteristics. Finally, it outputs results including inter-group difference values, significance determination, and visualization charts, solving the problem of insufficient scientific statistical support for multi-factor behavioral data comparisons. By adapting statistical testing methods to different data characteristics, the method provides quantification and significance determination of inter-group differences (e.g., analyzing the interaction effect between environment and genotype using linear models), providing rigorous statistical evidence for the study of the mechanisms by which space environments (such as microgravity and submagnetic fields) affect animal behavior, and enhancing the scientific rigor and persuasiveness of research conclusions.

[0102] To facilitate understanding of this solution, the following uses nematodes and zebrafish as examples to demonstrate the practical application of the analysis system 100 in four areas: basic movement behavior, special movement patterns, higher-order semantic behavior, and multi-factor comparison. All interfaces are driven by a unified JSON-frst output (pose, tracking, and frame-by-frame labels), and overlay and visualization are completed through an interactive front-end. The text focuses on showcasing representative interfaces and trends to further explain the methods provided in this embodiment of the invention.

[0103] In one example, *Caenorhabditis elegans* (Caenorhabditis elegans for short) is a non-toxic, harmless, and independently surviving model organism characterized by its small size, simple structure, short growth cycle, and high reproductive capacity. Its cellular mechanisms in response to space radiation and microgravity are similar to those of humans and some mammals. Furthermore, it exhibits good tolerance in orbit, requires minimal experimental conditions, and is easily observed under microscopy and genetically manipulated. Therefore, *C. elegans* is widely used in space life science research to explore the effects of the space environment on processes such as growth, development, reproduction, locomotion, and aging.

[0104] The data analyzed in this example comes from the on-orbit and ground-based matching experiment video of nematodes in the space station's life ecology experimental cabinet. The study utilizes a keypoint-driven multi-level behavioral analysis platform for space-based model animals, employing integrated posture estimation and tracking methods to acquire keypoints and trajectories. A deep behavioral analysis module is used to statistically analyze the number of head swaying, tail swaying, and body undulations. Figure 8 As shown, Figure 8 This is a schematic diagram of the interface of a nematode posture estimation and behavior analysis platform provided in an embodiment of the present invention. The top left corner displays a video overlay of posture estimation and tracking results; the bottom left corner displays a statistical information table linked to frame numbers; and the right side displays curves showing the changes in body curvature and head-tail angles over time. As can be seen, the platform interface provided in this embodiment supports video result overlay, switching between individual and segment views, linkage between statistical tables and video, and the display of dynamic changes in head-tail angles and curvature using time-series curves, facilitating result verification and motion feature analysis.

[0105] To further compare the differences in motion behavior between ground and on-orbit environments, this study selected three indicators: head swings (10-tal_head_swings), tail swings (total_tail_swings), and body bending (total_bends). A two-way ANOVA was used, combined with daily significance tests, to plot time-series curves. Figure 9 As shown, Figure 9The graph shows the differences in behavioral indicators of nematodes over time in ground and orbital environments. The horizontal axis represents the number of experimental days (Day), and the vertical axis represents the mean ± standard error (Mean ± SEM) of the behavioral indicators. The blue curve represents the ground group, and the orange curve represents the orbital group. The daily significance test results are marked between the two curves: "*" indicates p < 0.05, "**" indicates p < 0.01, "***" indicates p < 0.001, and "'ns" indicates no significance. The results show that time is the main effect driving changes in nematode movement behavior. `totalhead_swings` fluctuated significantly over time (p < 0.001), the main environmental effect was not significant (p = 0.996), but the interaction between time and environment was nearly significant (p = 0.050), showing that the orbital group was slightly higher than the ground group in the early stages of the experiment (days 1-2), and then converged. The total_iail_swings did not show a significant overall temporal trend (p=0.100), with the in-orbit group slightly lower than the ground group (p=0.063), and showing a significant decrease on days 7, 11, and 13 (interaction effect p=0.017). The total_bends showed the strongest phased fluctuations over time (p<1x10⁻⁸), with no significant main environmental effect (p=0.084), but a significant interaction effect (p=0.007), with the in-orbit group lower than the ground group on days 1 and 7. The nematode movement behavior generally exhibited a dynamic change over time, while the effect of the space environment was mainly reflected in the decrease in movement during specific phases. This suggests that microgravity may accelerate or alter the evolutionary process of individual movement behavior by modulating phased movement characteristics.

[0106] In addition, see Figure 10 This invention also compared the behavioral differences of different strains of nematodes in terrestrial and orbital environments. In the wild-type N2 strain, there were no significant differences in head swaying (p=0.050), tail swaying (p=0.719), and surging frequency (p=0.172); while in the s-gene defective type (deficient in the anti-muscular atrophy gene), only tail swaying was significantly reduced in the orbital group (p<0.001), while head swaying (p=0.922) and surging frequency (p=0.945) showed no significant differences. The overall analysis results are as follows: Figure 11 As shown, there was no significant difference in head swaying (p=0.375) and surging frequency (p=0.519), but tail swaying was significantly lower in the on-orbit group than in the ground group (p=0.011). These results suggest that the influence of the space environment on motor behavior is mainly concentrated on tail movement control, and there are differences among different strains, with the Is strain being more sensitive to the space environment. This suggests an interaction between genetic background and space environment, which has important reference value for studying space-induced muscle movement disorders and anti-atrophy mechanisms.

[0107] In another example, this invention uses zebrafish as a model organism to analyze speed and disorientation behavior, as detailed below: Zebrafish are small tropical freshwater fish, characterized by their small size, rapid reproduction, and short development cycle. Their genome shares over 80% similarity with the human genome, earning them the nickname "water mice." This example, based on zebrafish experimental videos from the space station's life ecology experimental cabinet, extracts six key points (mouth, back, abdomen, and tail 1 / 2 / 3). The movement trajectory is depicted using a center-of-gravity velocity curve, combined with heat maps of active areas and velocity-time curves, to analyze the basic movement characteristics of zebrafish and further identify disorientation behaviors (such as circling in place and swimming upside down). This example corresponds to the general movement behavior analysis and special behavior pattern analysis (disorientation) process, with the analysis interface as shown below. Figure 12 As shown.

[0108] See Figure 12 The platform interface of the analysis system 100 integrates multi-dimensional visualization results. The central video window overlays the attitude estimation skeleton in real time, supporting frame-by-frame browsing and manual verification. The right side displays the zebrafish's movement trajectory and active area heatmap; the former reflects the individual's swimming path within the experimental chamber, while the latter reveals its spatial distribution preferences and aggregation areas. Below are statistical curves showing the speed change over time; blue bars represent the average speed per second, and red fitted curves depict the overall trend, facilitating the identification of phased changes in movement activity. The disorientation behavior recognition module uses different colors on the timeline to represent two disorientation behaviors: blue represents swimming upside down, and orange represents spinning in place. Clicking on different colored blocks displays the corresponding continuous skeleton overlay in the window below the timeline. The skeleton's color, from light to dark, represents the evolution over time, and different colored dots represent different key points. The yellow dots represent the abdomen, and the orange dots represent the back. The bottom left view shows the zebrafish swimming belly-up and back-down, corresponding to "inverted swimming." The bottom right view shows the zebrafish forming a circle, indicating "circling in place." This module supports rapid identification of disorientation and obstacle behaviors in zebrafish and provides interactive visualization for easy verification and further observation of behavioral characteristics. In summary, this interface includes multi-level analysis from basic movement to abnormal patterns. The overall interface not only supports intuitive comparison of multi-source data but also facilitates further human-computer interaction and expert verification.

[0109] The foregoing mainly describes the solutions of the embodiments of the present invention from a methodological perspective. It is understood that, in order to achieve the above-mentioned functions, the analysis system 100 includes at least one of the hardware structures and software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, in conjunction with the units and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of the present invention can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of the present invention.

[0110] In this embodiment of the invention, the analysis system 100 can be divided into functional units according to the above method example. For example, the analysis system 100 can be divided into functional units corresponding to various functions, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0111] For example, Figure 13 A schematic diagram of the hardware structure of an analysis system provided in an embodiment of the present invention is shown. The analysis system 100 includes: an instruction receiving module 110, used to receive a first instruction input by a user, the first instruction being used to instruct the identification and quantitative analysis of abnormal movement behavior of a space model animal in a microgravity environment, and the first instruction carrying the model animal type, experimental video data, and environmental configuration parameters; a data processing module 120, used to respond to the first instruction, decode and extract frames from the experimental video data, and generate key point temporal trajectory data based on the coordinates of key feature points of the model animal in each frame of the experimental video data; and a detection module 130, used to perform analysis operations based on a special movement pattern detection algorithm matching the model animal type, combined with the environmental configuration parameters and the key point temporal trajectory data, to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, the quantitative index being the calculated value of the characteristic parameter of the abnormal behavior.

[0112] It should be understood that specific descriptions of the above-mentioned optional methods can be found in the foregoing method embodiments, and will not be repeated here. Furthermore, explanations of any of the analysis systems 100 provided above, as well as descriptions of their beneficial effects, can be found in the corresponding method embodiments described above, and will not be repeated here.

[0113] This invention also provides a computer-readable storage medium storing at least one computer instruction, which is loaded and executed by a processor to implement the methods of the various embodiments described above. Explanations of the relevant content and descriptions of the beneficial effects of any of the computer-readable storage media provided above can be found in the corresponding embodiments described above, and will not be repeated here.

[0114] This invention also provides a chip. This chip integrates a control circuit for implementing the functions of the analysis system 100 described above, and one or more ports. Optionally, the functions supported by this chip are as described above, and will not be repeated here.

[0115] Those skilled in the art will understand that the program for implementing all or part of the steps of the above embodiments, which can be executed by a program instructing related hardware, can be stored in a computer-readable storage medium. The storage medium mentioned above can be a read-only memory, a random access memory, etc. The processing unit or processor mentioned above can be a central processing unit, a general-purpose processor, an application-specific integrated circuit (ASIC), a microprocessor (DSP), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof.

[0116] This invention also provides a computer program product containing instructions that, when executed on a computer, cause the computer to perform any of the methods described in the above embodiments. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this invention is generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions may be transmitted from one website, computer, analysis system, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium accessible to a computer or a data storage device such as an analysis system or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., SSD), etc.

[0117] It should be noted that the devices for storing computer instructions or computer programs provided in the embodiments of the present invention, such as, but not limited to, the aforementioned memory, computer-readable storage medium, and communication chip, are all non-transitory. Those skilled in the art should recognize that the functions described in the embodiments of the present invention in one or more of the above examples can be implemented using hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium accessible to general-purpose or special-purpose computers.

[0118] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A keypoint-driven method for multi-level behavioral analysis of spatial model animals, characterized in that, The method includes: The system receives a first instruction from the user, which is used to instruct the identification and quantitative analysis of abnormal movement behavior of space model animals in a microgravity environment. The first instruction carries the model animal type, experimental video data and environmental configuration parameters. In response to the first instruction, the experimental video data is decoded and frame-sampling is performed, and key point temporal trajectory data is generated based on the coordinates of the feature key points of the model animal in each frame of the experimental video data. Based on the special movement pattern detection algorithm for matching the model animal type, and combined with the environmental configuration parameters and key point time-series trajectory data, an analysis operation is performed to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, wherein the quantitative index is the calculated value of the characteristic parameter of the abnormal behavior. The model organisms include zebrafish, fruit flies, and nematodes. When the model organism is zebrafish, the abnormal behaviors include spinning in place or being upside down. When the model organism is fruit flies, the abnormal behaviors include deviations between velocity and posture, acceleration contrary to posture, or spinning in place. When the model organism is nematodes, the abnormal behaviors include head-tail asynchrony or disordered pulsation rhythm. When the model animal type is zebrafish, the special motion pattern detection algorithm based on the model animal type matching performs analysis and calculation in conjunction with the environmental configuration parameters and key point time-series trajectory data to obtain a first analysis result, including: Based on the key point time-series trajectory data, the coordinates of key points of the zebrafish's mouth, back, abdomen, and tail are extracted; The quantitative indicators corresponding to the zebrafish are determined based on the environmental configuration parameters and the coordinates of key points of the zebrafish's mouth, back, abdomen and tail. The quantitative indicators include the head and tail orientation angle, the moving average angular velocity and the average translational velocity. If the zebrafish's sliding average angular velocity is greater than the angular velocity threshold and its average translational velocity is less than the upper limit of its stationary velocity, then the abnormal behavior of the zebrafish is determined to be spinning in place. In zebrafish y Back,t Greater than y Abdomen,t In this case, the abnormal behavior of the zebrafish was determined to be dorsoventrally inverted. The formula for determining the head and tail orientation angle of a zebrafish in frame t is: θ t =arctan2(y Tail3,t −y Mouth,t ,x Tail3,t −x Mouth,t ) ； Among them, the coordinates of the key points of the zebrafish's mouth are ( x Mouth,t , y Mouth,t The coordinates of the key points on the back are ( x Back,t ， y Back,t The coordinates of the key points in the abdomen are ( x Abdomen,t , y Abdomen,t The coordinates of the key points at the tail are ( x Tail3,t , y Tail3,t ); The moving average angular velocity of the zebrafish in frame t The formula for determining it is: ； ； Where W is the length of the sliding window. The wrap() function is used to reduce the angle difference between the head and tail orientation angles to the specified value. ; The formula for determining the average translational velocity of the zebrafish in frame t is: ； Where, p i Let be the center point of the zebrafish in the i-th frame.

2. The method according to claim 1, characterized in that, When the model animal type is fruit fly, the special movement pattern detection algorithm based on the model animal type matching performs analysis and calculation in conjunction with the environmental configuration parameters and key point time-series trajectory data to obtain a first analysis result, including: Based on the key point time-series trajectory data, the coordinates of key points on the mouth, left wing, right wing, back, abdomen, and tail of the fruit fly were extracted; The quantitative indicators corresponding to the fruit fly are determined based on the environmental configuration parameters and the coordinates of key points on the mouth, left wing, right wing, back, abdomen and tail of the fruit fly. The quantitative indicators include the angle between velocity and attitude, the angle between acceleration and attitude, and the attitude rotation angle. If the angle between the speed and attitude of the fruit fly is greater than the direction deviation threshold, the magnitude of the speed vector is greater than the first magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be a deviation between speed and attitude direction. If the angle between the acceleration and attitude of the fruit fly is greater than the acceleration contradiction threshold, the magnitude of the acceleration vector is greater than the third magnitude threshold, and the magnitude of the attitude direction is greater than the second magnitude threshold, the abnormal behavior type of the fruit fly is determined to be acceleration and attitude direction contradiction. If the angle difference of the fruit fly's posture rotation angle is greater than the minimum rotation angular velocity threshold and the magnitude of the velocity vector is less than the near-zero velocity threshold, the abnormal behavior type of the fruit fly is determined to be in-situ rotation. The velocity and attitude angle of the fruit fly in frame t The formula for determining it is: ； Among them, v t d is the velocity vector of the t-th frame; t The orientation is determined based on the coordinates of key points on the mouth and tail of the fruit fly; ∈[0,π]; The acceleration and attitude angle of a fruit fly in frame t The formula for determining it is: ； ； Among them, a t For acceleration; The formula for determining the attitude rotation angle of the fruit fly in frame t is: ； in, Let represent the attitude direction along the y-axis. The x-axis represents the attitude direction.

3. The method according to claim 1, characterized in that, When the model organism is a nematode, the special movement pattern detection algorithm based on the model organism type matching performs analysis and calculations in conjunction with the environmental configuration parameters and key point time-series trajectory data to obtain a first analysis result, including: The coordinates of five key points of the nematode were extracted based on the time-series trajectory data of the key points. The quantitative indicators corresponding to the nematode are determined based on the environmental configuration parameters and the coordinates of the 5 key points of the nematode. The quantitative indicators include the head swaying frequency, the tail swaying frequency, and the body undulation angle. If the absolute value of the difference between the head swaying frequency and the tail swaying frequency is greater than the frequency threshold, the abnormal behavior type of the nematode is determined to be head-tail asynchrony. The duration and amplitude of each two adjacent swing cycles of the nematode are determined based on the body undulation angle of the nematode in consecutive frames. If the difference between the duration of any two adjacent swing cycles exceeds a preset duration threshold or the difference between the amplitudes exceeds a preset amplitude threshold, the abnormal behavior type of the nematode is determined to be pulsating rhythm disorder. The formulas for determining the head swaying angle, tail swaying angle, and body undulation angle of the nematode in frame t are as follows: ； ； ； ； in, The angle between the two vectors, Let K be the coordinates of the k-th keypoint in frame t. The formula for determining the head-wagging frequency or tail-wagging frequency of a nematode is: ； Where C is the effective number of swings of the head or tail, the effective swing angle is greater than or equal to the preset angle threshold, and the number of continuous frames is greater than the preset number of continuous frames, T is the total number of frames, and F is the frame rate.

4. The method according to any one of claims 1-3, characterized in that, The method further includes: The system receives a second instruction from the user, which is used to instruct the quantitative analysis of the low-order kinematic characteristics of the space model animal, and the second instruction carries experimental video data and environmental configuration parameters. In response to the second instruction, the experimental video data is decoded and frame-segmented to obtain the skeletal center coordinate sequence of the model animal; The second analytical results of the model animal are determined based on the skeletal center coordinate sequence of the model animal. The second analytical results include velocity, acceleration, local average velocity, global average velocity, and activity. The velocity v t The formula for determining it is: ； in, Let t be the coordinates of the skeleton center. This refers to the inter-frame time interval. The acceleration a t The formula for determining it is: ； The local average velocity The formula for determining it is: ； The overall average speed The formula for determining it is: ； in, For the set of motion frame indices, ; Represents the Euclidean norm; k is the half-width of the window; The activity The formula for determining it is: ； Where T is the total number of frames; For indicator functions; This is the speed threshold.

5. The method according to claim 4, characterized in that, When the model organism is fruit fly, the method further includes: The system receives a third instruction from the user, which is used to instruct the identification and statistical analysis of semantic-level behavioral features of fruit flies, and the instruction carries experimental video data. In response to the third instruction, the head center coordinate sequence and feature key point angle data of the fruit fly are extracted, and the inter-frame displacement and total length of the historical trajectory of the fruit fly are determined. The semantic-level behavior category corresponding to each frame image is determined based on the inter-frame displacement and the total length of the historical trajectory of the fruit fly. The semantic-level behavior categories include stationary behavior, crawling behavior, flight behavior, courtship behavior, and mating behavior. Specifically, the semantic-level behavior category is determined as stationary behavior when the inter-frame displacement of the fruit fly is less than the upper limit of the frame-level displacement and the total length of the historical trajectory is less than the upper limit of the cumulative path of the short window. The semantic-level behavior category is determined as crawling behavior when the inter-frame displacement of the fruit fly is greater than or equal to the upper limit of the frame-level displacement and the total length of the historical trajectory is greater than or equal to the upper limit of the cumulative path of the short window. The semantic-level behavior category is determined as flight behavior when the inter-frame displacement of the fruit fly is greater than or equal to the flight speed threshold. The semantic-level behavior category is determined as courtship behavior when the absolute value of the difference between the angles of the left and right wings relative to the trunk axis of the fruit fly is greater than the left and right wing difference threshold, and the maximum value of the angles of the left and right wings relative to the trunk axis is greater than the minimum angle threshold. The semantic-level behavior category is mating behavior when the distance between the centers of the abdomen or trunk of two fruit flies is less than the contact distance threshold, the orientation difference is less than the orientation difference threshold, and the number of consecutive frames is greater than the number of consecutive frames threshold.

6. The method according to claim 5, characterized in that, The method further includes: The system receives a fourth instruction from the user, which is used to instruct the analysis of differences in behavioral data from different environments or generations, and the instruction carries at least two sets of experimental data, comparison dimension parameters, and significance test types. In response to the fourth instruction, core quantitative indicators are extracted from each set of experimental data. These core quantitative indicators include abnormal movement behavior characteristics, low-order kinematic characteristics, and semantic-level behavioral characteristics of the model animals. If the number of comparison groups K=2 and the data satisfies normality and homogeneity of variance, the t-test is used to calculate the test statistic. If the number of comparison groups K=2 and the normality or homogeneity of variance is not satisfied, the Mann-Whitney U test is used to calculate the test statistic. If the number of comparison groups K>2, use one-way ANOVA to calculate the test statistic; If the interaction effect of the two factors is considered, a linear model is used to calculate the test statistic; The fourth analysis results, based on the test statistics, include inter-group difference values, significance determination results, and visual comparison charts.

7. A keypoint-driven spatial model animal multi-level behavior analysis system, characterized in that, The system, applied to the method of any one of claims 1-6, comprises: The instruction receiving module is used to receive a first instruction input by the user. The first instruction is used to instruct the identification and quantitative analysis of abnormal movement behavior of space model animals in a microgravity environment. The first instruction carries the model animal type, experimental video data and environmental configuration parameters. The data processing module is used to respond to the first instruction to decode and extract frames from the experimental video data, and generate key point temporal trajectory data based on the coordinates of key feature points of the model animal in each frame of the experimental video data. The detection module is used to perform analysis and calculation based on the special movement pattern detection algorithm matching the model animal type, combined with the environmental configuration parameters and key point time-series trajectory data, to obtain a first analysis result; the first analysis result includes the abnormal movement behavior type corresponding to the model animal and the corresponding quantitative index, wherein the quantitative index is the calculated value of the characteristic parameter of the abnormal behavior.

8. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the keypoint-driven spatial model animal multilevel behavior analysis method as described in any one of claims 1-6.

Images